Bio572 - plotting

Some "rules" for plotting data

Use the right graph type to convey what you want... try multiple versions
- should reflect the type of data you have and how they are related
  - e.g., don't connect points if they are not from the same "individuals"
- distributions should be histograms or curves
- almost never reasons for 3D or pie charts!
Think about what information or quantity you want to plot
- quantity? density? ratio?
- In general, use the units common to the field, used in your text, or easiest to understand
Emphasize the right comparisons (the ones you think are important)
- between groups?
  - color or shape/symbol or facets
  - remember that we are better at comparing heights along a horizontal line, position vertically
- among series? try separate graphs above and below (no double axes)
- to the null expectation? Show your expectation (e.g., 1:1 line for before and after comparisons)
- Note that we are better at vertical comparisons (e.g., the heights of lines in two side-by-side plots) than horizontal ones
Be careful with your axis scales... changing them changes your meaning
- include zero (or not) depending on what you want to do
  - Note: always include zero with bar chart
- log is often good for growth, wealth, processes
Be honest with your best-fit lines (they can illuminate and mislead)
Get rid of any extraneous junk that doesn't help make the comparison(s)
The graph should stand alone
- axis labels should be clear and include units
- legend labels should be obvious
- label points when it helps make your point
- Generally, titles are useless